Homelessness, substance abuse, mental illness, and HIV complicate each other in ways that are not well understood. The goal of the proposed project is to conduct an evaluation of service interventions for homeless, HIV-positive, substance using adults with co-occuring mental illness. Data for analysis were collected by a national multi-site evaluation of 21 demonstration projects serving 2120 homeless or unstably housed substance using adults, with high rates of psychiatric disorder. A secondary goal of the project is to improve methodological tools available for the analysis of observational studies of service interventions. Observational studies can be an important complement to randomized controlled trials as a basis for improved service planning, yet present serious analytical challenges. The proposed research will compare different statistical approaches to dealing with these challenges. Specific aims of the research are: 1.) To empirically define and describe housing and service integration models of care; 2) To compare the relative advantages of alternative approaches to defining models of care for use in outcome studies of housing based interventions. We hypothesize that an empirically based classification of models of care will provide a more analytically and programmatically useful framework for comparative analysis than theoretically defined a priori models. Cluster analysis and latent class analysis will be used to help identify significant clusters in the full set of project level data. 3.) To examine outcomes for clients served by different models of care. Outcomes considered will be: utilization and adherence to AOD, health, and mental health treatment; drug use behaviors; sexual risk behaviors; housing stability; health and mental health functioning. Multi-level modeling statistical techniques will be used. Different approaches to propensity score weighting and missing value imputation will be compared as a means to adjust for pre-existing differences among treatment groups, and incomplete data. 4.) To translate research findings into program and policy recommendations with the assistance of an Advisory Group comprised of researchers, clinicians, service providers, policy experts, and consumers. The proposed project will contribute both substantive and methodological findings to current public health debates about how best to serve homeless HIV positive substance users with co-occurring mental illness and improve outcomes for individuals and their communities. [unreadable] [unreadable] [unreadable] [unreadable]